Pulmonary infiltrates remain as a diagnostic and therapeutic challenge in cancer patients. In order to evaluate the etiology, diagnostic methods used, Intensive Care Unit admission and in-hospital mortality, we conducted an observational, prospective study which included all patients with cancer and recent pulmonary infiltrates admitted to the Instituto Alexander Fleming between August 2003 and March 2006. Diagnostic methods were categorized in sequential steps of complexity: 1st step: radiological pattern of the pulmonary infiltrates, blood and sputum cultures, serological tests and empirical treatment response; 2nd step: bronchoalveolar lavage (BAL), non bronchoscopic tracheal aspirate and mini-BAL; 3rd step: pulmonary or extrapulmonary biopsies. Pulmonary infiltrate etiology was classified as: infection, treatment complication, disease progression, cardiovascular or mixed. Diagnosis was classified as proved or probable. A total of 106 samples from 103 patients were included. The etiologies were infection in 61 cases, disease progression in 4, treatment complication in 6, cardiovascular in 6 and mixed in 7. Proved diagnosis was obtained in 33 cases and probable diagnosis in 51 while 22 cases could not be diagnosed. Nine of the 10 diagnoses of mycoses were in oncohematologic cases. Seventy cases did not go further than procedures included in the 1st step. Thirty two cases stopped after diagnostic procedures of the 2nd step and 4 required biopsies. Forty four cases required Intensive Care Unit admission. In-hospital mortality was 30.2%. In our study, infection was the most frequent etiology. Mycoses were more frequent in oncohematologic cases. A proved or probable diagnosis was obtained in 84 (79.2%) cases. In 53.7% of the cases only non-invasive diagnostic methods were required.